How an audience signal works
In Performance Max and Demand Gen you do not choose exactly who sees the ad. Instead you hand Google a description of the people most likely to buy — your customer list, website visitors, people searching particular terms, interests, life events, demographics — and the system treats it as a starting point. It looks for the pattern in that group, then goes looking for people who resemble it.
The important word is signal. Nothing in it stops Google serving outside the group. Once the campaign has learned enough from real conversions, its own data outweighs the description you supplied, and the signal fades into background information rather than a rule.
Why audience signals matter
They shorten the learning phase. A campaign starting cold has to discover who converts by spending money on people who do not, and a good signal narrows that search considerably. On a small budget in a small market — which describes most accounts in Nepal — that difference decides whether a campaign ever reaches a stable cost per enquiry.
The quality of the signal also sets the ceiling on what the system can learn. Feed it a list of people who genuinely paid you and it looks for more people like them. Feed it everyone who ever landed on the site and it will happily find more browsers.
Common mistakes with audience signals
Treating it as targeting is the first and most expensive. Advertisers add a tight signal, watch traffic arrive from outside it, and conclude the campaign is broken. It is behaving exactly as designed; if you need a hard boundary, that is a different setting or a different campaign type.
Next is a vague signal — a broad interest category or a demographic covering half the country. That teaches the system nothing it did not already assume. Then there is a signal built on the wrong conversion. If the account counts every form fill as success, including the ones that never turn into work, the signal quietly trains the campaign to find more of the wrong enquiry.
How to act on it
Build the signal from the strongest evidence you own before reaching for anything Google offers. A customer match list of people who actually paid comes first, then buyers or high-intent visitors from your own site, then search terms you know produce work, then interests and demographics last. Combine inputs that describe the same person rather than stacking unrelated ideas on top of each other.
Then leave it alone long enough to learn, and judge it on the quality of the enquiries rather than on where the impressions landed. If the campaign drifts towards cheap, low-value conversions, fix the conversion definition and the customer data feeding Performance Max before touching the signal itself.